I was just reading through the mlogit section of the reference manual.
On page 514-515 there is a section on setting up a classification table
following a mlogit run.
On page 515 there is a table showing the classification table that was
generated.
the commentary on this table states that "There are a substantial number of
misclassifications in both directions, though there are more correctly
classified observations than misclassified observations".
My reading of the table is that 134 out of 616 (21.8%) are correctly
classified.
Am I missing something here?
Kieran
The wording is a little puzzling. Perhaps they are excluding the
388 "ambiguous" cases from the calculations, in which case 134 of the
remaining 228 cases are correctly classified.
